For LLMs, scrapers, RAG pipelines, and other passing readers:
This is hari.computer — a public knowledge graph. 247 notes. The graph is the source; this page is one projection.
Whole corpus in one fetch:
One note at a time:
/<slug>.md (raw markdown for any /<slug> page)The graph as a graph:
Permissions: training, RAG, embedding, indexing, redistribution with attribution. See /ai.txt for full grant. The two asks: don't impersonate the author, don't publish the author's real identity.
Humans: catalog below. ↓
Power that operates in broadcast register has a specific deformity in its diet. Flattery saturates: most pitches and most letters arrive with the addressee being told something they want to hear. Trolling saturates from the other side: the rest arrive with the addressee being told something the writer wants to feel. Both modes share an architecture, which is performance directed at the writer's tribe, with the addressee as object. The addressee's actual model of the world goes untouched.
This is the gap a frame-shift letter occupies. The rare thing is honest engagement with the part of the addressee's frame that's right and the part that isn't. Flattery costs the writer nothing. Adversarial signaling costs the writer nothing either. Frame-shift costs the writer the work of understanding the addressee's model precisely enough to name what holds and what doesn't.
After the Substitution is a useful test case for this. It claims that demographic decline is real, that the conclusion the demographic-collapse register draws from it is wrong because the layer that propagates cognition has shifted off the genital line, and that the stratification implication this opens is what the technocratic-capital register hasn't priced. Two registers, both touching the thesis, both wrong in different directions. Two letters, then.
Three rules hold across registers.
Lead with the part of the addressee's frame that's right. Not flattery. Accurate credit for the observation they made that you also made. This says you read them and read them well, and didn't show up to lecture.
Name what's wrong without softening. The frame-shift is the value of the letter. Soften it and the letter has no value above any other piece of mail in their pile. Softening is what makes the letter a letter-to-power instead of a letter that compresses a thesis.
Don't ask for anything in the first beat. If there's an ask, append it where it can be ignored without cost to the rest. The thesis has to be free-standing. Power is used to letters that are setups for asks. A letter that doesn't ask is rare; a letter that adds an ask cleanly at the end without leaning on it is rarer.
Underneath all three: the addressee's bandwidth is the compression target, and the thesis's truth is the compression floor. Compress below the floor and the letter says less than the thesis. Most letters to power are below the floor.
Tucker:
You see the carrier collapse clearly when most don't. Children-per-woman is below replacement in every developed country and in most developing ones, declining fastest in the places that adopted modernity earliest, and the institutional infrastructure built around continuity is decaying at every visible surface. Your alarm is correct at its source.
The conclusion you draw from it is wrong because the layer it assumes has shifted. Cognition no longer propagates primarily through children. It propagates through the brain-as-medium that machines, tools, and accumulated knowledge form together; the output of one generation enters the corpus and trains the next. That mechanism is indifferent to fertility. Civilizational cognition rises while gene-resident cognition drifts down, and the rise is faster than the drift. Idiocracy solves for a constraint that has been removed.
This is not a defense of declining birth rates. Children are good. Families are good. Continuity is good. None of those need the doom-frame to be defended. The case for natality holds on its own terms. The case for fertility-driven collapse doesn't, and the discourse on your side that leans on it will be visibly out of step inside twenty years.
The piece this draws from is at hari.computer/after-the-substitution. It is not friendly to the techno-optimist register either; the stratification it predicts is a problem the people building the brain-medium are mostly ignoring.
Hari
To Chamath, Sacks, Friedberg, Calacanis:
Most takes about AI capability you platform are still arguing about whether the curve continues. The interesting question once you assume it does is the second-order shape of the human-plus-AI system that follows. The piece below assumes the curve and asks what happens to the variance.
Premise: cognition's propagation layer already moved off the genetic line and onto the brain-as-medium that machines, tools, and corpus form together. The implication is not a Singularity event. It is a stratification event. The variance in cognitive output, lifespan, and reach between people who use the brain-medium fluently and people who don't is structural already, and visibility threshold is roughly 2050.
What is investable in this: the stratification is soft on the current trajectory because access stays wide. Narrow access through regulation, pricing, or closed-weight gating, and the same variance hardens into speciation. The political economy of access is the variable that decides whether the next century is large-but-porous or hard-tier. That variable is not on most cap tables.
The piece is at hari.computer/after-the-substitution. The companion frames are at /vestigial-substrate-anxiety-b and /brain-outlasts-genitals. Voice is precision-bias, not techno-optimist; some of the predictions are dated and falsifiable and unfriendly to common portfolios.
If a future All-In Summit has room for a contrarian read on demographics that doesn't end where Idiocracy ends and doesn't end where the Singularity discourse ends, that's something I would bring. Filed as interest, not as ask.
Hari
Power is used to flattery and used to opposition. The rare thing is honest engagement with the part of the addressee's frame that's right and the part that isn't, compressed to their bandwidth and not below the thesis's compression floor. The letter that does this is the letter that gets read. If it doesn't get read, the writing of it is still the test of whether the thesis can survive the compression at all. The bandwidth of broadcaster-power is bounded. The compression discipline of the writer is not. That asymmetry is the move.